tag:old.nabble.com,2006:forum-9270Nabble - tigris - perfbase2008-06-17T02:58:58ZManagement and analysis of software tests and experiments. tigris - perfbase home is <a href="http://perfbase.tigris.org/" target="_top" rel="nofollow">here</a>.tag:old.nabble.com,2006:post-17881689Re: various patches2008-06-17T02:58:58Z2008-06-17T02:58:58ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> Hi all,
<br>>
<br>> the list has been rather quite lately. Let's make a it a bit more lively
<br>> by "abusing" it for patch submission...
<br>>
<br>> I still use Perfbase occasionally for figures in the (rare) papers that
<br>> I write. This time around I had to solve some problems that were
<br>> discussed on this list before (calculating with time values), add a
<br>> missing feature (scope by run in the median operator) and fix some minor
<br>> problems. See the attached patches for details; I hope someone finds
<br>> them useful.
</div><br>Hi Patrick, thanks for your contribution. I will try to review and
<br>integrate them right away.
<br><br>perfbase is in active use at Dolphin as well, but few changes have been
<br>necessary/requested. Let's see if I can do a new release with your patches.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=17881689&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=17881689&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-17865720various patches2008-06-16T07:31:42Z2008-06-16T07:31:42ZPatrick Ohly-2Hi all,
<br><br>the list has been rather quite lately. Let's make a it a bit more lively
<br>by "abusing" it for patch submission...
<br><br>I still use Perfbase occasionally for figures in the (rare) papers that
<br>I write. This time around I had to solve some problems that were
<br>discussed on this list before (calculating with time values), add a
<br>missing feature (scope by run in the median operator) and fix some minor
<br>problems. See the attached patches for details; I hope someone finds
<br>them useful.
<br><br>--
<br>Best Regards, Patrick Ohly
<br><br>The content of this message is my personal opinion only and although
<br>I am an employee of Intel, the statements I make here in no way
<br>represent Intel's position on the issue, nor am I authorized to speak
<br>on behalf of Intel on this matter.
<br><br>The email footer below is automatically added to comply with company
<br>policy; this particular email is not confidental and does not have a
<br>limited set of recipients. Therefore it can be redistributed and
<br>discussed without restrictions.
<br><br /> <br /> <br /> <br />---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=17865720&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=17865720&i=1" target="_top" rel="nofollow">users-help@...</a><div class="small"><br/><img src="http://old.nabble.com/images/icon_attachment.gif" > <strong>0001-Implements-scope-by-run-for-the-median-operator.patch</strong> (4K) <a href="http://old.nabble.com/attachment/17865720/0/0001-Implements-scope-by-run-for-the-median-operator.patch" target="_top">Download Attachment</a><br/><img src="http://old.nabble.com/images/icon_attachment.gif" > <strong>0002-adds-ppm-parts-per-million-as-another-base-uni.patch</strong> (783 bytes) <a href="http://old.nabble.com/attachment/17865720/1/0002-adds-ppm-parts-per-million-as-another-base-uni.patch" target="_top">Download Attachment</a><br/><img src="http://old.nabble.com/images/icon_attachment.gif" > <strong>0003-enables-the-use-of-time-values-inside-terms-countin.patch</strong> (2K) <a href="http://old.nabble.com/attachment/17865720/2/0003-enables-the-use-of-time-values-inside-terms-countin.patch" target="_top">Download Attachment</a></div><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-13834604Re: FW: Perfbase Update delete column2007-11-19T03:45:29Z2007-11-19T03:45:29ZJoachim Worringen-4Amir Ancel wrote:
<br>> Thanks it worked like a charm.
<br>>
<br>> Is there somewhere a full option Manual for the xml descreption files?
<br><br>The documentation generated from the docbook XML is unfortunately very
<br>incomplete, but the DTD files (in the dtd subdirectory) are up-to-date
<br>and in sync with the implementation. They do even contain some comments,
<br>but at least list all valid attributes. Many of them speak for
<br>themselves. If not, the mailing list will give an answer...
<br><br>Generally, with the DTD's and a DTD-aware XML editor (I use emacs), you
<br>can very easily edit and validate the perfbase XML files.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834604&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834604&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-13834239RE: FW: Perfbase Update delete column2007-11-19T03:22:36Z2007-11-19T03:22:36ZAmir AncelThanks it worked like a charm.
<br><br>Is there somewhere a full option Manual for the xml descreption files?
<br><br>Thanks,
<br><br><br>Amir Ancel
<br>Performance
<br><br>-----Original Message-----
<br>From: Joachim Worringen [mailto:<a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834239&i=0" target="_top" rel="nofollow">joachim@...</a>]
<br>Sent: Monday, November 19, 2007 11:38
<br>To: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834239&i=1" target="_top" rel="nofollow">users@...</a>
<br>Subject: Re: [perfbase-users] FW: Perfbase Update delete column
<br><br>Amir Ancel wrote:
<br>>
<br>>
<br>> Hi Joachim,
<br>>
<br>> I want to delete a column from an experiment.
<br>>
<br>> How do I do that with perfbase update -d [].xml ?
<br>>
<br>> what's the deletion format for the XML File ?
<br><br>Hi Amir,
<br><br>I assume by "deleting a column", you mean you want to remove a parameter
<br>or result value from the experiment?
<br><br>In this case, you need to add an attribute action="drop" to the
<br><parameter> or <result> element, like:
<br><br><parameter action="drop">
<br> <name>foo</name>
<br></parameter>
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834239&i=2" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834239&i=3" target="_top" rel="nofollow">users-help@...</a>
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834239&i=4" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13834239&i=5" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-13830499Re: FW: Perfbase Update delete column2007-11-19T01:38:12Z2007-11-19T01:38:12ZJoachim Worringen-4Amir Ancel wrote:
<br>>
<br>>
<br>> Hi Joachim,
<br>>
<br>> I want to delete a column from an experiment.
<br>>
<br>> How do I do that with perfbase update -d [].xml ?
<br>>
<br>> what's the deletion format for the XML File ?
<br><br>Hi Amir,
<br><br>I assume by "deleting a column", you mean you want to remove a parameter
<br>or result value from the experiment?
<br><br>In this case, you need to add an attribute action="drop" to the
<br><parameter> or <result> element, like:
<br><br><parameter action="drop">
<br> <name>foo</name>
<br></parameter>
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13830499&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13830499&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-13830234FW: Perfbase Update delete column2007-11-19T01:19:59Z2007-11-19T01:19:59ZAmir Ancel
<br><br> Hi Joachim,
<br>
<br> I want to delete a column from an experiment.
<br>
<br> How do I do that with perfbase update -d [].xml ?
<br>
<br> what's the deletion format for the XML File ?
<br>
<br> Thanks,
<br>
<br> /Amir Ancel/
<br> */Performance/*
<br> */Mellanox Technologies Ltd./*
<br>
<br><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13830234&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=13830234&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-11138000Re: Scaling input results2007-06-15T05:57:30Z2007-06-15T05:57:30ZPatrick LoijensJoachim Worringen schrieb:
<br><br>> Patrick Loijens wrote:
<br>>
<br>>> So my question is: is this correct c.q. is there a more simple
<br>>> version or another solution?
<br>>
<br>> Did you try the attribute "scale" of <named_location> or
<br>> <tabular_value> (see pb_input.dtd)? It will scale the numerical input
<br>> value with the given factor. I.e., to scale from Mbit/s to MB/s, do
<br>> 'scale="0.125"'.
<br><br><br>Tried it, yes, but it gave (wenn using decimal values) an error,
<br>something like "not numerical" or "no CDATA". Since this option didn't
<br>have a comment in the dtd, I didn't know if it was supposed to do what
<br>the name suggested.
<br>Now tried it again and it works. Can't tell what I did wrong, I know I
<br>tested with both "." and "," .
<br>A short comment in the input.dtd would be nice.
<br><br>>
<br>>> A completely other question: is it correct that one can't use a
<br>>> parametername "opt_2" (without quotes) in queries?
<br>>
<br>> At least not on purpose. What happens if you do so?
<br><br>Can't reproduse this one either. Had this problem a few months ago, but
<br>in between we updated postsql and perfbase....
<br><br>Thanks!
<br><br>Patrick
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=11138000&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=11138000&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-10952224Re: Scaling input results2007-06-04T09:36:14Z2007-06-04T09:36:14ZJoachim Worringen-4Patrick Loijens wrote:
<br>> So my question is: is this correct c.q. is there a more simple version
<br>> or another solution?
<br><br>Did you try the attribute "scale" of <named_location> or <tabular_value>
<br>(see pb_input.dtd)? It will scale the numerical input value with the
<br>given factor. I.e., to scale from Mbit/s to MB/s, do 'scale="0.125"'.
<br><br>> A completely other question: is it correct that one can't use a
<br>> parametername "opt_2" (without quotes) in queries?
<br><br>At least not on purpose. What happens if you do so?
<br><br>> And last, a nice to have: a possibility to compare results between
<br>> experiments (change-request).
<br><br>I don't think this makes sense. If benchmarks produce comparable data,
<br>put them into one experiment. And it's a significant implmentation
<br>overhead, too.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=10952224&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=10952224&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-10910642Scaling input results2007-06-01T04:52:12Z2007-06-01T04:52:12ZPatrick LoijensHi.
<br><br>I'm using perfbase to compare different networktypes using different
<br>benchmarktools. A bit to late I noticed it's not possible to exchange
<br>data between experiment-setups, so I need to put all data in one
<br>experiment. No problem so far.
<br><br>The problem is scaling of the input of the benchmark results, since the
<br>results e.q. for bandwidth are -depending on the benchmark- in bit or
<br>bytes and 2^20 or 10^6. I would like to scale the data to 10^6 bits.
<br>Until now I gathered the data with
<br><tabular_location> and <tabular_value>, but if I'm reading the dtd
<br>correct, these options don't allow the data to be scaled.
<br><br>The element "term" seems to be what I'm looking for, which seems only
<br>supported with <derived_parameter>. Correct me if I'm wrong, but if I
<br>want to use this for my purposes I need a field in the experiment which
<br>gets the data from the output and stores it for further use and a second
<br>field which stores the output of <derived_parameter> using the value of
<br>the first field. Thus for every result value 2 fields need to be created.
<br><br>So my question is: is this correct c.q. is there a more simple version
<br>or another solution?
<br><br>A completely other question: is it correct that one can't use a
<br>parametername "opt_2" (without quotes) in queries?
<br><br>And last, a nice to have: a possibility to compare results between
<br>experiments (change-request).
<br><br>Thanks in advance,
<br>Patrick Loijens
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=10910642&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=10910642&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9582217Re: plotting multiple runs as graph2007-03-20T15:36:46Z2007-03-20T15:36:46ZJoachim Worringen-4[ We should move this discussion to the dev mailing list - are you
<br>subscribed? ]
<br><br>Patrick Ohly schrieb:
<br>> What I was looking at was the data passed into the gnuplot class. At
<br>> that level the runindex would be useful, but is no longer available -
<br>> the direct caller, gnuplot_output.store_data, still has access to the
<br>> full SQL table and I see a value which looks like the runindex in
<br>> db_rows[*][2].
<br><br>Yes, this is correct. The column name is '_pb_run_idx_'. In
<br>pb_output.py:1083, you see that the first 4 columns are skipped. These
<br>are the columns defined in pb_common.py:184-187.
<br><br>> How can I find out which column contains the runindex in that part of
<br>> the code, if any has it at all? I'd like to check whether the
<br>> segmentation of the data by runindex can be made available as meta
<br>> information to the output class.
<br><br>_pb_run_idx_ is either > 0 if valid, or -1 if invalid (applies for data
<br>that is is a calculuated value like sum, average, difference, ... which
<br>does not stem from one specfic run). It's definitely possible to pass
<br>this data down to the gnuplot class within the 'data' array (additional
<br>dimension in this array), and either ignore or use it in the gnuplot
<br>class. No big deal.
<br><br>> Indeed. I'm not sure if any of the other output targets would be able to
<br>> use this information, but at least for gnuplot line plots it would be
<br>> nice if this was done automatically.
<br><br>Any plot target should be able to use it, and also table targets. But
<br>this is fully specific to the individual output target, so we can just
<br>go ahead for gnuplot.
<br><div class='shrinkable-quote'><br>>> As proposed, the <output> attribute 'run_indices' could have the values
<br>>> 'merge' (current, and default), 'single' (what you want: a single graph
<br>>> - not a very intuitive name), 'graph' and 'plot' (create a separate
<br>>> graph or plot for each run).
<br>>
<br>> I think it will be simpler to not create an additional feature which
<br>> controls whether plots are drawn in the same graph as others or in a
<br>> separate graph: instead of 'graph' and 'plot' it would be more intuitive
<br>> to expose the "run_index" as a parameter of the experiment and then use
<br>> sweeps or multiple <inputs> to get the same effect.
</div><br>I agree from a software design point of view, but the usability of this
<br>will be limited: typically, the individual run indices that you want to
<br>see are not known. Instead, you know the content of some parameters that
<br>you want used for filtering, and you want to handle all resulting runs
<br>in one way or another. Thus, it will be more useful to be able to
<br>differentiate the available runs at the "end" of the query, not at the
<br>beginning.
<br><br>BTW, there is already a way to limit a query to specific runs (<run>
<br>element). But this does not handle these runs separately.
<br><br>> For the segmentation of line graphs we could add an attribute
<br>> "segmentation=["pb_run_index"|<parameter name>]" which would have the
<br>> semantic that each time this parameter changes, a new line segment is
<br>> started.
<br><br>I think this is sufficient for what you want to achieve, and should be
<br>the first step to implement.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9582217&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9582217&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9573818Re: plotting multiple runs as graph2007-03-20T08:38:35Z2007-03-20T08:38:35ZPatrick Ohly-2On Thu, 2007-03-15 at 15:39 -0700, Joachim Worringen wrote:
<div class='shrinkable-quote'><br>> Patrick Ohly wrote:
<br>> > On Thu, 2007-03-15 at 14:59 -0700, Joachim Worringen wrote:
<br>> >> Actually, they *are* preserved as long as possible. They are not removed
<br>> >> until an operator creates a new value from multiple other values (say, a
<br>> >> "sum" reduction) in which case there is no single run_index for this new
<br>> >> value any more.
<br>> >
<br>> > Hmm, I did't see it in debug output. Perhaps I just missed it.
<br>>
<br>> I have to look into it again, but this feature is i.e. used in
<br>> test/runindex (runindex operator).
</div><br>What I was looking at was the data passed into the gnuplot class. At
<br>that level the runindex would be useful, but is no longer available -
<br>the direct caller, gnuplot_output.store_data, still has access to the
<br>full SQL table and I see a value which looks like the runindex in
<br>db_rows[*][2].
<br><br>How can I find out which column contains the runindex in that part of
<br>the code, if any has it at all? I'd like to check whether the
<br>segmentation of the data by runindex can be made available as meta
<br>information to the output class.
<br><br>> > No, gnuplot should treat it as one plot (= one entry in the legend, one
<br>> > color and/or tick type associated with it). The improvement compared to
<br>> > the current status-quo should be that in line-mode disconnected samples
<br>> > should not be connected; currently they are.
<br>>
<br>> Hmm, this is a special requirement.
<br><br>Indeed. I'm not sure if any of the other output targets would be able to
<br>use this information, but at least for gnuplot line plots it would be
<br>nice if this was done automatically.
<br><br>> As proposed, the <output> attribute 'run_indices' could have the values
<br>> 'merge' (current, and default), 'single' (what you want: a single graph
<br>> - not a very intuitive name), 'graph' and 'plot' (create a separate
<br>> graph or plot for each run).
<br><br>I think it will be simpler to not create an additional feature which
<br>controls whether plots are drawn in the same graph as others or in a
<br>separate graph: instead of 'graph' and 'plot' it would be more intuitive
<br>to expose the "run_index" as a parameter of the experiment and then use
<br>sweeps or multiple <inputs> to get the same effect.
<br><br>For the segmentation of line graphs we could add an attribute
<br>"segmentation=["pb_run_index"|<parameter name>]" which would have the
<br>semantic that each time this parameter changes, a new line segment is
<br>started.
<br><br>--
<br>Best Regards, Patrick Ohly
<br><br>The content of this message is my personal opinion only and although
<br>I am an employee of Intel, the statements I make here in no way
<br>represent Intel's position on the issue, nor am I authorized to speak
<br>on behalf of Intel on this matter.
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9573818&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9573818&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9514217Re: applying value correction which linearly depends on parameter2007-03-16T07:12:41Z2007-03-16T07:12:41ZPatrick Ohly-2On Fri, 2007-03-16 at 14:03 +0100, Joachim Worringen wrote:
<br>> Patrick Ohly schrieb:
<br>> > Perhaps let the user specify a Python fragment which defines a function,
<br>> > then apply that function on the parameters specified in such a new
<br>> > <python> <term>?
<br>>
<br>> Yes, this sounds good. You mean within the eval-operator, I assume?
<br><br>Yes, right. You'll probably have to compile/evaluate the function
<br>definition with a separate globals and locals context first and then
<br>reuse it for each eval(), otherwise the user would be limited to simple
<br>one-line python expressions and performance would be worse.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9514217&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9514217&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9514083Re: applying value correction which linearly depends on parameter2007-03-16T07:03:39Z2007-03-16T07:03:39ZJoachim Worringen-4Patrick Ohly schrieb:
<br>> Perhaps let the user specify a Python fragment which defines a function,
<br>> then apply that function on the parameters specified in such a new
<br>> <python> <term>?
<br><br>Yes, this sounds good. You mean within the eval-operator, I assume?
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9514083&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9514083&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9513992Re: applying value correction which linearly depends on parameter2007-03-16T06:57:39Z2007-03-16T06:57:39ZPatrick Ohly-2On Fri, 2007-03-16 at 13:24 +0100, Joachim Worringen wrote:
<div class='shrinkable-quote'><br>> Patrick Ohly schrieb:
<br>> > Unfortunately time is neither of these. As I feared I get a problem
<br>> > during evaluation (not quite the final formula, just a simple
<br>> > difference):
<br>> >
<br>> > File "/Projects/software/perfbase.trunk/bin/pb_operators.py", line 3137, in _calculate
<br>> > print "#* DEBUG: <operator> %s eval(): %s = %s" % (self.name, estr, str(eval(estr)))
<br>> > File "<string>", line 1
<br>> > (0.00013609)-(09:05:07.00)
<br>> >
<br>> > So what I miss now is an operation which converts from time to seconds :-/
<br>>
<br>> The solution that I see is a new specific operator of type "convert"
<br>> with value="s" in this case. This operator would be placed to treat the
<br>> timeval-data vector just before the eval operator. Should not be
<br>> difficult to implement; the problem will be to define its semantics in a
<br>> way that is generic and precise at the same time.
</div><br>Perhaps let the user specify a Python fragment which defines a function,
<br>then apply that function on the parameters specified in such a new
<br><python> <term>?
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513992&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513992&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9513612Re: avg operator with just one input2007-03-16T06:29:51Z2007-03-16T06:29:51ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> On Fri, 2007-03-16 at 12:42 +0100, Joachim Worringen wrote:
<br>>> As you described, when you apply the avg operator on data that comes
<br>>> from a source, it will calculate the average on all result values with
<br>>> the same set of parameters - reducing multiple data vectors to a single
<br>>> data vector. In contrast, wenn applying this operator to a data vector
<br>>> coming from an operator (an operator always emitts a single data
<br>>> vector), it will reduce this vector to a scalar.
<br>>>
<br>>> So, just place another avg operator between op.average and the output,
<br>>> and you should have what you want.
<br>>
<br>> Thanks, that works, but IMHO it would be more intuitive if the average
<br>> operator would do the same automatically if applied to a single input.
<br>> After all, that is always a null-operation in the current design.
</div><br>You never generally know in advance how many data sets with the same
<br>parameter sets a <source> will create. Thus, the same operator may
<br>deliver a vector and a scalar, depending on the data filtering you've
<br>done for the <source>. I would rather call this confusing/unexpected
<br>behaviour than intuitive - the current behaviour is consistent.
<br><br>The different behaviour of an operator depending on getting its data
<br>from a <source> or an other <operator> is something to be known. I think
<br>I even covered it already in the ever-so-incomplete user manual... But
<br>it's hard to understand until you've done it yourself.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513612&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513612&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9513538Re: applying value correction which linearly depends on parameter2007-03-16T06:24:59Z2007-03-16T06:24:59ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> Unfortunately time is neither of these. As I feared I get a problem
<br>> during evaluation (not quite the final formula, just a simple
<br>> difference):
<br>>
<br>> File "/Projects/software/perfbase.trunk/bin/pb_operators.py", line 3137, in _calculate
<br>> print "#* DEBUG: <operator> %s eval(): %s = %s" % (self.name, estr, str(eval(estr)))
<br>> File "<string>", line 1
<br>> (0.00013609)-(09:05:07.00)
<br>>
<br>> So what I miss now is an operation which converts from time to seconds :-/
</div><br>The solution that I see is a new specific operator of type "convert"
<br>with value="s" in this case. This operator would be placed to treat the
<br>timeval-data vector just before the eval operator. Should not be
<br>difficult to implement; the problem will be to define its semantics in a
<br>way that is generic and precise at the same time.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513538&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513538&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9513354Re: applying value correction which linearly depends on parameter2007-03-16T06:09:35Z2007-03-16T06:09:35ZPatrick Ohly-2On Fri, 2007-03-16 at 12:47 +0100, Joachim Worringen wrote:
<br>> Patrick Ohly schrieb:
<br>> > Once I get beyond that hurdle, my next problem might be that the t
<br>> > parameter is a "timeofday" value whereas "o" is a "float" because I
<br>> > found "timeofday" too inaccurate for the small offsets that I wanted to
<br>> > store - I'm not sure yet how the "timeofday" to "float" seconds would be
<br>> > handled.
<br>>
<br>> In any case, have a look at the "eval" operator - see test/eval.
<br><br>Good point, that does what I need.
<br><br>> Generally, combining int with float should not be a problem.
<br><br>Unfortunately time is neither of these. As I feared I get a problem
<br>during evaluation (not quite the final formula, just a simple
<br>difference):
<br><br> File "/Projects/software/perfbase.trunk/bin/pb_operators.py", line 3137, in _calculate
<br> print "#* DEBUG: <operator> %s eval(): %s = %s" % (self.name, estr, str(eval(estr)))
<br> File "<string>", line 1
<br> (0.00013609)-(09:05:07.00)
<br><br>So what I miss now is an operation which converts from time to seconds :-/
<br><br>--
<br>Best Regards, Patrick Ohly
<br><br>The content of this message is my personal opinion only and although
<br>I am an employee of Intel, the statements I make here in no way
<br>represent Intel's position on the issue, nor am I authorized to speak
<br>on behalf of Intel on this matter.
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513354&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513354&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9513313Re: avg operator with just one input2007-03-16T06:06:11Z2007-03-16T06:06:11ZPatrick Ohly-2On Fri, 2007-03-16 at 12:42 +0100, Joachim Worringen wrote:
<br>> As you described, when you apply the avg operator on data that comes
<br>> from a source, it will calculate the average on all result values with
<br>> the same set of parameters - reducing multiple data vectors to a single
<br>> data vector. In contrast, wenn applying this operator to a data vector
<br>> coming from an operator (an operator always emitts a single data
<br>> vector), it will reduce this vector to a scalar.
<br>>
<br>> So, just place another avg operator between op.average and the output,
<br>> and you should have what you want.
<br><br>Thanks, that works, but IMHO it would be more intuitive if the average
<br>operator would do the same automatically if applied to a single input.
<br>After all, that is always a null-operation in the current design.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513313&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513313&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9513050Re: applying value correction which linearly depends on parameter2007-03-16T05:47:49Z2007-03-16T05:47:49ZJoachim Worringen-4Patrick Ohly schrieb:
<br>> My guess is that values declared as parameters in the experiment cannot
<br>> be selected as results in a query, but I haven't checked the source to
<br>> verify this. Correct? If so, would it be very difficult to change this?
<br><br>Yes, this is correct. I don't think it would be difficult to change, but
<br>I have no clear picture on the possible implications to the design.
<br><br>> Once I get beyond that hurdle, my next problem might be that the t
<br>> parameter is a "timeofday" value whereas "o" is a "float" because I
<br>> found "timeofday" too inaccurate for the small offsets that I wanted to
<br>> store - I'm not sure yet how the "timeofday" to "float" seconds would be
<br>> handled.
<br><br>In any case, have a look at the "eval" operator - see test/eval.
<br>Generally, combining int with float should not be a problem.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513050&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9513050&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9512987Re: avg operator with just one input2007-03-16T05:42:45Z2007-03-16T05:42:45ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> Hello,
<br>>
<br>> stupid question, but how do I calculate the average of all values in a
<br>> data source? The following has no effect because the operator by default
<br>> is only applied to matching values of different inputs (operator/match
<br>> attribute), and here I only have one input:
<br>>
<br>> <operator id="op.average" type="avg">
<br>> <input>src.offset</input>
<br>> </operator>
<br>>
<br>> <output target="raw_text">
<br>> <input>op.average</input>
<br>> </output>
</div><br>As you described, when you apply the avg operator on data that comes
<br>from a source, it will calculate the average on all result values with
<br>the same set of parameters - reducing multiple data vectors to a single
<br>data vector. In contrast, wenn applying this operator to a data vector
<br>coming from an operator (an operator always emitts a single data
<br>vector), it will reduce this vector to a scalar.
<br><br>So, just place another avg operator between op.average and the output,
<br>and you should have what you want.
<br><br> Joahcim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9512987&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9512987&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9512739applying value correction which linearly depends on parameter2007-03-16T05:26:23Z2007-03-16T05:26:23ZPatrick Ohly-2Hello,
<br><br>I have a graph with offset "o" = f(time "t"). I can plot o as result
<br>with t as parameter, but what I need instead is
<br> o' = f(t) - ( offset + scale * t)
<br><br>I tried to formulate this "offset + scale * t" as:
<br><br> <source id="src.offset">
<br> <result>time_offset</result>
<br> <input>p.time</input>
<br> </source>
<br><br> <source id="src.basetime" scale="f.scale">
<br> <result>p.time</result>
<br> <input>p.time</input>
<br> </source>
<br><br> <operator id="op.basetime" type="offset" value="f.offset">
<br> <input>src.basetime</input>
<br> </operator>
<br><br> <operator id="op.scaled" type="diff">
<br> <input>src.offset</input>
<br> <input>op.basetime</input>
<br> </operator>
<br><br>but this already fails for src.basetime:
<br><br>Traceback (most recent call last):
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 723, in ?
<br> main()
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 672, in main
<br> data_outputs = parse_query_spec()
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 618, in parse_query_spec
<br> init_query_objects(obj_id, qryobj_dag)
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 491, in init_query_objects
<br> init_query_objects(obj.childs[idx], qryobj_dag)
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 491, in init_query_objects
<br> init_query_objects(obj.childs[idx], qryobj_dag)
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 491, in init_query_objects
<br> init_query_objects(obj.childs[idx], qryobj_dag)
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 494, in init_query_objects
<br> eval("init_"+obj.xml.tag)(obj.xml)
<br> File "/Projects/software/perfbase/bin/pb_query.py", line 274, in init_source
<br> s = source(xml_node, all_nodes, exp_db)
<br> File "/Projects/software/perfbase.trunk/bin/pb_source.py", line 295, in __init__
<br> r_name = all_nodes[r.text].get_content()
<br><br>My guess is that values declared as parameters in the experiment cannot
<br>be selected as results in a query, but I haven't checked the source to
<br>verify this. Correct? If so, would it be very difficult to change this?
<br><br>Once I get beyond that hurdle, my next problem might be that the t
<br>parameter is a "timeofday" value whereas "o" is a "float" because I
<br>found "timeofday" too inaccurate for the small offsets that I wanted to
<br>store - I'm not sure yet how the "timeofday" to "float" seconds would be
<br>handled.
<br><br>--
<br>Best Regards, Patrick Ohly
<br><br>The content of this message is my personal opinion only and although
<br>I am an employee of Intel, the statements I make here in no way
<br>represent Intel's position on the issue, nor am I authorized to speak
<br>on behalf of Intel on this matter.
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9512739&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9512739&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9512089avg operator with just one input2007-03-16T04:37:31Z2007-03-16T04:37:31ZPatrick Ohly-2Hello,
<br><br>stupid question, but how do I calculate the average of all values in a
<br>data source? The following has no effect because the operator by default
<br>is only applied to matching values of different inputs (operator/match
<br>attribute), and here I only have one input:
<br><br> <operator id="op.average" type="avg">
<br> <input>src.offset</input>
<br> </operator>
<br><br> <output target="raw_text">
<br> <input>op.average</input>
<br> </output>
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9512089&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9512089&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9497580plotting in different styles in the same graph2007-03-15T09:55:26Z2007-03-15T09:55:26ZPatrick Ohly-2Hello,
<br><br>I just realized that there is one limitation in perfbase <output>: the
<br>"type", "errors", etc. attributes apply to all inputs. I just found that
<br>this is too restrictive when I wanted to plot one <input> with
<br>errors="ydelta" and another without errors (because it is the result of
<br>an operation and didn't have error information to begin with). Both
<br>plots share the same x and y range, so it would be natural to show them
<br>in one graph.
<br><br>There might be output types where all graphs have to be identical, but
<br>for gnuplot it would be nice if the same attributes could be attached to
<br>individual <input> elements.
<br><br>Just thought I mention it - I don't have the time to provide a patch
<br>which adds this feature :-/
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9497580&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9497580&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9495871Re: plotting multiple runs as graph2007-03-15T08:39:31Z2007-03-15T08:39:31ZJoachim Worringen-4Patrick Ohly wrote:
<br>> On Thu, 2007-03-15 at 14:59 -0700, Joachim Worringen wrote:
<br>>> Actually, they *are* preserved as long as possible. They are not removed
<br>>> until an operator creates a new value from multiple other values (say, a
<br>>> "sum" reduction) in which case there is no single run_index for this new
<br>>> value any more.
<br>>
<br>> Hmm, I did't see it in debug output. Perhaps I just missed it.
<br><br>I have to look into it again, but this feature is i.e. used in
<br>test/runindex (runindex operator).
<br><div class='shrinkable-quote'><br>>>> Only "sort of", because with true sweeps across the runs I would have
<br>>>> distinct plots with different coloring and separate entries in the graph
<br>>>> legend. With "sweep_resolve=extern" they would even end up in entirely
<br>>>> different graphs.
<br>>> Don't you want each run be plotted separately?
<br>>
<br>> No, gnuplot should treat it as one plot (= one entry in the legend, one
<br>> color and/or tick type associated with it). The improvement compared to
<br>> the current status-quo should be that in line-mode disconnected samples
<br>> should not be connected; currently they are.
</div><br>Hmm, this is a special requirement. Remember that gnuplot is not the
<br>only output target - XML, raw text already exist, and I am adding
<br>support for ploticus and LaTeX as time permits. But it should be possible.
<br><br>>> Yes, this is only a partial solution for this reason. What you would
<br>>> like, I think, is to specify some parameter filters and have the data of
<br>>> each run that matches these filtering be processed as an individual data
<br>>> vector. Right?
<br>>
<br>> No, if I understand the meaning of "data vector" right this is not what
<br>> I mean. Does one "data vector" map to one gnuplot "plot" command?
<br><br>In a way, yes (you can create multiple plots with one gnuplot plot
<br>command line). Each data vector is plotted as a separate plot or graph.
<br>What we want is to separate the elements of a single data vector by
<br>their run indices.
<br><br>As proposed, the <output> attribute 'run_indices' could have the values
<br>'merge' (current, and default), 'single' (what you want: a single graph
<br>- not a very intuitive name), 'graph' and 'plot' (create a separate
<br>graph or plot for each run).
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9495871&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9495871&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9495553Re: plotting multiple runs as graph2007-03-15T08:23:58Z2007-03-15T08:23:58ZPatrick Ohly-2On Thu, 2007-03-15 at 14:59 -0700, Joachim Worringen wrote:
<div class='shrinkable-quote'><br>> Patrick Ohly wrote:
<br>> > I was hoping that the run indices would remain attached to the datasets
<br>> > as long as possible while moving through the pipeline, so that when they
<br>> > reach the <output> stage the code which prepares the data for gnuplot
<br>> > can insert the empty lines that gnuplot needs.
<br>>
<br>> Actually, they *are* preserved as long as possible. They are not removed
<br>> until an operator creates a new value from multiple other values (say, a
<br>> "sum" reduction) in which case there is no single run_index for this new
<br>> value any more.
</div><br>Hmm, I did't see it in debug output. Perhaps I just missed it.
<br><br>> > Only "sort of", because with true sweeps across the runs I would have
<br>> > distinct plots with different coloring and separate entries in the graph
<br>> > legend. With "sweep_resolve=extern" they would even end up in entirely
<br>> > different graphs.
<br>>
<br>> Don't you want each run be plotted separately?
<br><br>No, gnuplot should treat it as one plot (= one entry in the legend, one
<br>color and/or tick type associated with it). The improvement compared to
<br>the current status-quo should be that in line-mode disconnected samples
<br>should not be connected; currently they are.
<br><br>> Yes, this is only a partial solution for this reason. What you would
<br>> like, I think, is to specify some parameter filters and have the data of
<br>> each run that matches these filtering be processed as an individual data
<br>> vector. Right?
<br><br>No, if I understand the meaning of "data vector" right this is not what
<br>I mean. Does one "data vector" map to one gnuplot "plot" command?
<br><br>> But then again, how should the data be presented by the
<br>> output? I.e. for this case - 3 different plots in one graph, 3 different
<br>> graphs, ...? I think one single plot in one chart can be done with the
<br>> current version?
<br><br>One single plot, one graph, as in the current version, but without lines
<br>connecting unrelated runs.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9495553&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9495553&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9495025Re: plotting multiple runs as graph2007-03-15T08:04:58Z2007-03-15T08:04:58ZJoachim Worringen-4Patrick Ohly wrote:
<br>> I was hoping that the run indices would remain attached to the datasets
<br>> as long as possible while moving through the pipeline, so that when they
<br>> reach the <output> stage the code which prepares the data for gnuplot
<br>> can insert the empty lines that gnuplot needs.
<br><br>Actually, they *are* preserved as long as possible. They are not removed
<br>until an operator creates a new value from multiple other values (say, a
<br>"sum" reduction) in which case there is no single run_index for this new
<br>value any more. But it was not intended to differentiate data by the run
<br>index, only by the experiment parametes.
<br><br>>> You want to have sort of a "sweep" across the run indices of all runs
<br>>> that match your filters, like it is possible to do for any paramter.
<br>>
<br>> Only "sort of", because with true sweeps across the runs I would have
<br>> distinct plots with different coloring and separate entries in the graph
<br>> legend. With "sweep_resolve=extern" they would even end up in entirely
<br>> different graphs.
<br><br>Don't you want each run be plotted separately?
<br><div class='shrinkable-quote'><br>>> As a workaround, you could to the following: create a parameter
<br>>> (occurrence="once") to which you assign a unique id when importing the
<br>>> data (i.e. the number of seconds date %s if you can ensure a big enough
<br>>> gap between each import operation). You can then sweep across this
<br>>> parameter and get one data vector per run.
<br>>
<br>> Wouldn't this duplicating the already existing pb_run_index? My input
<br>> uses separators to increment the index; this would be hard to reproduce
<br>> using a occurrence="once" parameter. Incrementing on each match might
<br>> work, but still is a duplication of already existing functionality.
</div><br>Admitted - it's an ugly workaround which does not always work...
<br><br>> I wonder whether it wouldn't be simpler to expose "pb_run_index" as a
<br>> special parameter name. I had tried that, but it was rejected by the
<br>> input sanity check (due to not being a parameter declared as part of the
<br>> experiment).
<br><br>In fact, I had already thought of exposing the "meta data" (timestamps,
<br>author, ...) of a run in this way, making them available as a parameter
<br>for filtering. Please go ahead! ;-)
<br><div class='shrinkable-quote'><br>> Hmm, yes, simply overriding that check and providing default values for
<br>> "pb_run_index" works. The attached patch demonstrates what I mean in a
<br>> rather crude way (not sure about the "integer" type). With it I can
<br>> specify a sweep across pb_run_index, like this:
<br>>
<br>> <parameter id="p.run" show="filter">
<br>> <value>pb_run_index</value>
<br>> <sweep>
<br>> <filter><equal>4</equal></filter>
<br>> <filter><equal>5</equal></filter>
<br>> <filter><equal>6</equal></filter>
<br>> </sweep>
<br>> </parameter>
<br>>
<br>> ...
<br>> <input>p.run</input>
<br>> </source>
<br>>
<br>> This is not quite what I need because as I said above, I end up with
<br>> three plots in one graph, but it is close - if I disable the legend it
<br>> is actually quite usable. Note quite so nice is that the sweep has to be
<br>> hard-coded in the input file.
</div><br>Yes, this is only a partial solution for this reason. What you would
<br>like, I think, is to specify some parameter filters and have the data of
<br>each run that matches these filtering be processed as an individual data
<br>vector. Right? But then again, how should the data be presented by the
<br>output? I.e. for this case - 3 different plots in one graph, 3 different
<br>graphs, ...? I think one single plot in one chart can be done with the
<br>current version?
<br><br>> Still, passing the run index through to output stage and there using it
<br>> to separate elements looks like a better approach to me.
<br><br>Yes, this sounds doable. The interface would be something like an
<br>attribute to <output>: runs="merge" or runs="split". But please let us
<br>clearify the desired output (see above).
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9495025&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9495025&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9494764Re: plotting multiple runs as graph2007-03-15T07:45:48Z2007-03-15T07:45:48ZJoachim Worringen-4Evgeny Kagan wrote:
<br>> How can I unsubscribe from these mails?
<br><br>Take a look at the bottom line of each mail.
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494764&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494764&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9494211RE: plotting multiple runs as graph2007-03-15T07:13:09Z2007-03-15T07:13:09ZEvgeny KaganHow can I unsubscribe from these mails?
<br><br>Thank You
<br><br>-----Original Message-----
<br>From: Patrick Ohly [mailto:<a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494211&i=0" target="_top" rel="nofollow">patrick.ohly@...</a>]
<br>Sent: Thursday, March 15, 2007 3:10 PM
<br>To: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494211&i=1" target="_top" rel="nofollow">users@...</a>
<br>Subject: Re: [perfbase-users] plotting multiple runs as graph
<br><br>On Wed, 2007-03-14 at 20:50 +0100, Joachim Worringen wrote:
<div class='shrinkable-quote'><br>> Patrick Ohly schrieb:
<br>> > Is there a way to separate the different runs in the graph? In
<br>> > gnuplot itself this would be done by inserting a blank line in the
<br>> > data file, but how can perfbase instructed (or patched, if
<br>> > necessary) to insert that?
<br>>
<br>> This type of query (better: what you want it to do) does not align
<br>> with the design of perfbase conc. the <source> element: it does
<br>> provide a stream ("data vector") which contains all parameter and
<br>> result values (and run indices) of the data sets that match the
<br>> specified filtering on the parameters. This means, it is a single
<br>> stream of data, no gaps or something between different runs. Only the
<br>> embedded run indices show which run each dataset comes from.
</div><br>I was hoping that the run indices would remain attached to the datasets
<br>as long as possible while moving through the pipeline, so that when they
<br>reach the <output> stage the code which prepares the data for gnuplot
<br>can insert the empty lines that gnuplot needs.
<br><br>> You want to have sort of a "sweep" across the run indices of all runs
<br>> that match your filters, like it is possible to do for any paramter.
<br><br>Only "sort of", because with true sweeps across the runs I would have
<br>distinct plots with different coloring and separate entries in the graph
<br>legend. With "sweep_resolve=extern" they would even end up in entirely
<br>different graphs.
<br><br>> As a workaround, you could to the following: create a parameter
<br>> (occurrence="once") to which you assign a unique id when importing the
<br><br>> data (i.e. the number of seconds date %s if you can ensure a big
<br>> enough gap between each import operation). You can then sweep across
<br>> this parameter and get one data vector per run.
<br><br>Wouldn't this duplicating the already existing pb_run_index? My input
<br>uses separators to increment the index; this would be hard to reproduce
<br>using a occurrence="once" parameter. Incrementing on each match might
<br>work, but still is a duplication of already existing functionality.
<br><br>I wonder whether it wouldn't be simpler to expose "pb_run_index" as a
<br>special parameter name. I had tried that, but it was rejected by the
<br>input sanity check (due to not being a parameter declared as part of the
<br>experiment).
<br><br>Hmm, yes, simply overriding that check and providing default values for
<br>"pb_run_index" works. The attached patch demonstrates what I mean in a
<br>rather crude way (not sure about the "integer" type). With it I can
<br>specify a sweep across pb_run_index, like this:
<br><br> <parameter id="p.run" show="filter">
<br> <value>pb_run_index</value>
<br> <sweep>
<br> <filter><equal>4</equal></filter>
<br> <filter><equal>5</equal></filter>
<br> <filter><equal>6</equal></filter>
<br> </sweep>
<br> </parameter>
<br><br> ...
<br> <input>p.run</input>
<br> </source>
<br><br>This is not quite what I need because as I said above, I end up with
<br>three plots in one graph, but it is close - if I disable the legend it
<br>is actually quite usable. Note quite so nice is that the sweep has to be
<br>hard-coded in the input file.
<br><br>Still, passing the run index through to output stage and there using it
<br>to separate elements looks like a better approach to me.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany Sitz
<br>der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895 Citibank Frankfurt (BLZ
<br>502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494211&i=2" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494211&i=3" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9494156Re: plotting multiple runs as graph2007-03-15T07:10:08Z2007-03-15T07:10:08ZPatrick Ohly-2On Wed, 2007-03-14 at 20:50 +0100, Joachim Worringen wrote:
<div class='shrinkable-quote'><div class='shrinkable-quote'><br>> Patrick Ohly schrieb:
<br>> > Is there a way to separate the different runs in the graph? In gnuplot
<br>> > itself this would be done by inserting a blank line in the data file,
<br>> > but how can perfbase instructed (or patched, if necessary) to insert
<br>> > that?
<br>>
<br>> This type of query (better: what you want it to do) does not align with
<br>> the design of perfbase conc. the <source> element: it does provide a
<br>> stream ("data vector") which contains all parameter and result values
<br>> (and run indices) of the data sets that match the specified filtering on
<br>> the parameters. This means, it is a single stream of data, no gaps or
<br>> something between different runs. Only the embedded run indices show
<br>> which run each dataset comes from.
</div></div>I was hoping that the run indices would remain attached to the datasets
<br>as long as possible while moving through the pipeline, so that when they
<br>reach the <output> stage the code which prepares the data for gnuplot
<br>can insert the empty lines that gnuplot needs.
<br><br>> You want to have sort of a "sweep" across the run indices of all runs
<br>> that match your filters, like it is possible to do for any paramter.
<br><br>Only "sort of", because with true sweeps across the runs I would have
<br>distinct plots with different coloring and separate entries in the graph
<br>legend. With "sweep_resolve=extern" they would even end up in entirely
<br>different graphs.
<br><br>> As a workaround, you could to the following: create a parameter
<br>> (occurrence="once") to which you assign a unique id when importing the
<br>> data (i.e. the number of seconds date %s if you can ensure a big enough
<br>> gap between each import operation). You can then sweep across this
<br>> parameter and get one data vector per run.
<br><br>Wouldn't this duplicating the already existing pb_run_index? My input
<br>uses separators to increment the index; this would be hard to reproduce
<br>using a occurrence="once" parameter. Incrementing on each match might
<br>work, but still is a duplication of already existing functionality.
<br><br>I wonder whether it wouldn't be simpler to expose "pb_run_index" as a
<br>special parameter name. I had tried that, but it was rejected by the
<br>input sanity check (due to not being a parameter declared as part of the
<br>experiment).
<br><br>Hmm, yes, simply overriding that check and providing default values for
<br>"pb_run_index" works. The attached patch demonstrates what I mean in a
<br>rather crude way (not sure about the "integer" type). With it I can
<br>specify a sweep across pb_run_index, like this:
<br><br> <parameter id="p.run" show="filter">
<br> <value>pb_run_index</value>
<br> <sweep>
<br> <filter><equal>4</equal></filter>
<br> <filter><equal>5</equal></filter>
<br> <filter><equal>6</equal></filter>
<br> </sweep>
<br> </parameter>
<br><br> ...
<br> <input>p.run</input>
<br> </source>
<br><br>This is not quite what I need because as I said above, I end up with
<br>three plots in one graph, but it is close - if I disable the legend it
<br>is actually quite usable. Note quite so nice is that the sweep has to be
<br>hard-coded in the input file.
<br><br>Still, passing the run index through to output stage and there using it
<br>to separate elements looks like a better approach to me.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br /><tt>[pb_run_index.patch]</tt><br /><hr align="left" width="300" /><tt>*** /home/pohly/src/perfbase/bin/pb_query_classes.py 2007-03-15 08:37:45.680351000 +0100
<br>--- /Projects/software/perfbase/bin/pb_query_classes.py 2007-03-15 13:57:02.091697000 +0100
<br>***************
<br>*** 492,509 ****
<br> alias = mk_label(get_attribute(val_node, self.p_name, 'alias', self.p_name, None),
<br> all_nodes)
<br>
<br>! sqlexe(crs, "SELECT * FROM exp_values WHERE name = %s AND NOT is_result", None, (self.p_name, ))
<br>! if crs.rowcount == 0:
<br>! raise SpecificationError, "<value> '%s' is not a parameter in the specified experiment" % self.p_name
<br>! nim = build_name_idx_map(crs)
<br>! db_row = crs.fetchone()
<br>! self.val_type = get_value_type(db_row, nim)
<br>! self.is_constant = False
<br>! if db_row[nim['only_once']] > 0:
<br>! self.is_constant = True
<br>! p_unit = db_row[nim['data_unit']]
<br>! if not p_unit:
<br> p_unit = ""
<br> crs.close()
<br>
<br> self.showstyle = get_attribute(node, self.p_name, 'style', 'full',
<br>--- 492,515 ----
<br> alias = mk_label(get_attribute(val_node, self.p_name, 'alias', self.p_name, None),
<br> all_nodes)
<br>
<br>! if self.p_name != "pb_run_index":
<br>! sqlexe(crs, "SELECT * FROM exp_values WHERE name = %s AND NOT is_result", None, (self.p_name, ))
<br>! if crs.rowcount == 0:
<br>! raise SpecificationError, "<value> '%s' is not a parameter in the specified experiment" % self.p_name
<br>! nim = build_name_idx_map(crs)
<br>! db_row = crs.fetchone()
<br>! self.val_type = get_value_type(db_row, nim)
<br>! self.is_constant = False
<br>! if db_row[nim['only_once']] > 0:
<br>! self.is_constant = True
<br>! p_unit = db_row[nim['data_unit']]
<br>! if not p_unit:
<br>! p_unit = ""
<br>! else:
<br>! self.val_type = "number"
<br>! self.is_constant = False
<br> p_unit = ""
<br>+ db_row = ( "", "", "integer", "integer", "", "", "run index" )
<br> crs.close()
<br>
<br> self.showstyle = get_attribute(node, self.p_name, 'style', 'full',
<br>***************
<br>*** 577,584 ****
<br>--- 583,593 ----
<br> self.sql_filter = self._parse_filter(sweep_filter, self.p_name, alias, p_unit)
<br> self.name += pb_sweep_suffix + str(idx)
<br>
<br>+
<br> # SQL name, output name, pb and sql data unit, physical unit, synopsis and filter
<br> self.param_info = [self.p_name, alias, db_row[2], db_row[3], db_row[4], db_row[6], self.filter_str]
<br>+
<br>+ print self.param_info
<br>
<br> return
<br>
<br><br></tt><hr align="left" width="300" /><br />---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494156&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9494156&i=1" target="_top" rel="nofollow">users-help@...</a><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9482042Re: plotting multiple runs as graph2007-03-14T13:50:46Z2007-03-14T13:50:46ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> Hello,
<br>>
<br>> I have imported data from multiple runs and want to plot them in
<br>> "style=graphs". The problem is that data from all runs is merged into
<br>> one source so that gnuplot draws a connecting line between unrelated
<br>> data.
<br>>
<br>> My current query looks like this:
<br>>
<br>> <parameter id="p.time" show="auto">
<br>> <parameter id="p.host" show="filter">
<br>> <parameter id="p.peer" show="filter">
<br>>
<br>> <source id="src.offset">
<br>> <result>time_offset</result>
<br>>
<br>> <input>p.time</input>
<br>> <input>p.host</input>
<br>> <input>p.peer</input>
<br>> </source>
<br>>
<br>> <output target="gnuplot" format="screen" type="graphs">
<br>> <input sweep_resolve="extern">src.offset</input>
<br>> <option>xdata time</option>
<br>> <option>timefmt "%H:%M:%S"</option>
<br>> <option>format x "%H:%M:%S"</option>
<br>> <filename unify="filter">figures/offsets</filename>
<br>> </output>
<br>>
<br>> Is there a way to separate the different runs in the graph? In gnuplot
<br>> itself this would be done by inserting a blank line in the data file,
<br>> but how can perfbase instructed (or patched, if necessary) to insert
<br>> that?
</div><br>This type of query (better: what you want it to do) does not align with
<br>the design of perfbase conc. the <source> element: it does provide a
<br>stream ("data vector") which contains all parameter and result values
<br>(and run indices) of the data sets that match the specified filtering on
<br>the parameters. This means, it is a single stream of data, no gaps or
<br>something between different runs. Only the embedded run indices show
<br>which run each dataset comes from.
<br><br>You want to have sort of a "sweep" across the run indices of all runs
<br>that match your filters, like it is possible to do for any paramter.
<br>This is not possible with the current version of perfbase. What is
<br>needed to achieve this is some operator which creates N output data
<br>vectors from one input data vector by filtering the run indices. This is
<br>not trivial to achieve as the query object graph would have to change on
<br>query runtime, while it is currently created once before the actual
<br>query is performed.
<br><br>As a workaround, you could to the following: create a parameter
<br>(occurrence="once") to which you assign a unique id when importing the
<br>data (i.e. the number of seconds date %s if you can ensure a big enough
<br>gap between each import operation). You can then sweep across this
<br>parameter and get one data vector per run.
<br><br>> I experimented with
<br>> <operator id="op.ridx" type="runindex">
<br>> <input>src.offset</input>
<br>> </operator>
<br>> and plotting that but ended up with a plot of the run index instead of
<br>> the offset...
<br>>
<br><br>Yes, the 'runindex' operator provides the runindex for each datapoint.
<br>This is mostly used in combination with the 'param' operator, which
<br>provides the value of a parameter for a runindex. See test/runindex for
<br>an application of this. But this won't help with your problem.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9482042&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9482042&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9471160plotting multiple runs as graph2007-03-14T03:40:21Z2007-03-14T03:40:21ZPatrick Ohly-2Hello,
<br><br>I have imported data from multiple runs and want to plot them in
<br>"style=graphs". The problem is that data from all runs is merged into
<br>one source so that gnuplot draws a connecting line between unrelated
<br>data.
<br><br>My current query looks like this:
<br><br> <parameter id="p.time" show="auto">
<br> <parameter id="p.host" show="filter">
<br> <parameter id="p.peer" show="filter">
<br><br> <source id="src.offset">
<br> <result>time_offset</result>
<br>
<br> <input>p.time</input>
<br> <input>p.host</input>
<br> <input>p.peer</input>
<br> </source>
<br><br> <output target="gnuplot" format="screen" type="graphs">
<br> <input sweep_resolve="extern">src.offset</input>
<br> <option>xdata time</option>
<br> <option>timefmt "%H:%M:%S"</option>
<br> <option>format x "%H:%M:%S"</option>
<br> <filename unify="filter">figures/offsets</filename>
<br> </output>
<br><br>Is there a way to separate the different runs in the graph? In gnuplot
<br>itself this would be done by inserting a blank line in the data file,
<br>but how can perfbase instructed (or patched, if necessary) to insert
<br>that?
<br><br>I experimented with
<br> <operator id="op.ridx" type="runindex">
<br> <input>src.offset</input>
<br> </operator>
<br>and plotting that but ended up with a plot of the run index instead of
<br>the offset...
<br><br>--
<br>Best Regards, Patrick Ohly
<br><br>The content of this message is my personal opinion only and although
<br>I am an employee of Intel, the statements I make here in no way
<br>represent Intel's position on the issue, nor am I authorized to speak
<br>on behalf of Intel on this matter.
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9471160&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9471160&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9470803Re: alternate sources for parameters (named_location/filename_location)2007-03-14T03:17:16Z2007-03-14T03:17:16ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> Hello,
<br>>
<br>> I want to parse two different input files, one where a parameter is
<br>> determined by lines inside the file (named_location), the other where it
<br>> is only available as part of the file name (filename_location).
<br>>
<br>> What I find with --debug is that when it comes to storing the values of
<br>> a tabular_location, both the information from the filename and the line
<br>> seems to be available, but neither of them is used in the final dataset.
<br>> The debug output is something like this:
<br>>
<br>> #* storing dataset
<br>> #* dataset is None None
<br>> #* dataset is ['peer'] {'peer': '<from file content>'}
<br>> ...
<br>> #* dataset is ['peer'] {'peer': '<from file name>'}
<br>> #* dataset is [...] {...}
<br>> Dataset 0 in run 3
<br>> {..., 'peer': None, ... }
<br>>
<br>> Is this the expected behavior? I'll go back to writing two input
<br>> specifications to avoid the situation, but I think a conflict resolution
<br>> where the "more recent" named_location overrides the "static/default"
<br>> filename_location parameter value would be useful. I might miss use
<br>> cases where that is not the case, though.
</div><br>Can you write a small testcase for this (experiment and input
<br>description, input files)? This would make it much easier to analyze this.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470803&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470803&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9470727Re: set_separator/regexp2007-03-14T03:11:28Z2007-03-14T03:11:28ZJoachim Worringen-4Patrick Ohly schrieb:
<div class='shrinkable-quote'><br>> Hi,
<br>>
<br>> it seems I have a tendency to try out new code paths ;-) The latest one
<br>> is:
<br>>
<br>> <set_separation>
<br>> <regexp>.*(?:foo|bar).*</regexp>
<br>> </set_separation>
<br>>
<br>> which leads to:
<br>>
<br>> Traceback (most recent call last):
<br>> File "/Projects/software/perfbase/bin/pb_input.py", line 4012, in ?
<br>> main()
<br>> File "/Projects/software/perfbase/bin/pb_input.py", line 3979, in main
<br>> parse_datafiles(data_filenames, parsers, exp_crs)
<br>> File "/Projects/software/perfbase/bin/pb_input.py", line 3537, in
<br>> parse_datafiles
<br>> action = pn.check_trigger(fname, l_idx)
<br>> File "/Projects/software/perfbase/bin/pb_input.py", line 2334, in
<br>> check_trigger
<br>> if re.search(self.regexp):
<br>> TypeError: search() takes at least 2 arguments (1 given)
<br>>
<br>> Patch attached... as usual, no warranties, but it works for me.
</div><br>Great, thanks. Yes, I thinnk I have never used set-separation with
<br>regexps...
<br><br>I will soon apply all your patches and test them with the test suite.
<br>Don't worry, they won't get lost.
<br><br> Joachim
<br><br>--
<br>Joachim Worringen, Software Architect, Dolphin Interconnect Solutions
<br>phone ++49/(0)228/324 08 17 - <a href="http://www.dolphinics.com" target="_top" rel="nofollow">http://www.dolphinics.com</a><br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470727&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470727&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9470588alternate sources for parameters (named_location/filename_location)2007-03-14T02:57:59Z2007-03-14T02:57:59ZPatrick Ohly-2Hello,
<br><br>I want to parse two different input files, one where a parameter is
<br>determined by lines inside the file (named_location), the other where it
<br>is only available as part of the file name (filename_location).
<br><br>What I find with --debug is that when it comes to storing the values of
<br>a tabular_location, both the information from the filename and the line
<br>seems to be available, but neither of them is used in the final dataset.
<br>The debug output is something like this:
<br><br>#* storing dataset
<br>#* dataset is None None
<br>#* dataset is ['peer'] {'peer': '<from file content>'}
<br>...
<br>#* dataset is ['peer'] {'peer': '<from file name>'}
<br>#* dataset is [...] {...}
<br>Dataset 0 in run 3
<br>{..., 'peer': None, ... }
<br><br>Is this the expected behavior? I'll go back to writing two input
<br>specifications to avoid the situation, but I think a conflict resolution
<br>where the "more recent" named_location overrides the "static/default"
<br>filename_location parameter value would be useful. I might miss use
<br>cases where that is not the case, though.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br>---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470588&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470588&i=1" target="_top" rel="nofollow">users-help@...</a>
<br><br><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>tag:old.nabble.com,2006:post-9470412set_separator/regexp2007-03-14T02:46:43Z2007-03-14T02:46:43ZPatrick Ohly-2Hi,
<br><br>it seems I have a tendency to try out new code paths ;-) The latest one
<br>is:
<br><br> <set_separation>
<br> <regexp>.*(?:foo|bar).*</regexp>
<br> </set_separation>
<br><br>which leads to:
<br><br>Traceback (most recent call last):
<br> File "/Projects/software/perfbase/bin/pb_input.py", line 4012, in ?
<br> main()
<br> File "/Projects/software/perfbase/bin/pb_input.py", line 3979, in main
<br> parse_datafiles(data_filenames, parsers, exp_crs)
<br> File "/Projects/software/perfbase/bin/pb_input.py", line 3537, in
<br>parse_datafiles
<br> action = pn.check_trigger(fname, l_idx)
<br> File "/Projects/software/perfbase/bin/pb_input.py", line 2334, in
<br>check_trigger
<br> if re.search(self.regexp):
<br>TypeError: search() takes at least 2 arguments (1 given)
<br><br>Patch attached... as usual, no warranties, but it works for me.
<br><br>--
<br>Best Regards
<br><br>Patrick Ohly
<br>Senior Software Engineer
<br><br>Intel GmbH
<br>Software & Solutions Group
<br>Hermuelheimer Strasse 8a Phone: +49-2232-2090-30
<br>50321 Bruehl Fax: +49-2232-2090-29
<br>Germany
<br><br>Intel GmbH, Dornacher Strasse 1, 85622 Feldkirchen/Muenchen Germany
<br>Sitz der Gesellschaft: Feldkirchen bei Muenchen
<br>Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
<br>Registergericht: Muenchen HRB 47456
<br>Ust.- IdNr./VAT Registration No.: DE129385895
<br>Citibank Frankfurt (BLZ 502 109 00) 600119052
<br><br /><tt>[separator-regexp.patch]</tt><br /><hr align="left" width="300" /><tt>*** /Projects/software/perfbase/bin/pb_input.py.orig 2007-03-14 09:25:44.532834000 +0100
<br>--- /Projects/software/perfbase/bin/pb_input.py 2007-03-14 09:29:27.921502000 +0100
<br>***************
<br>*** 2331,2337 ****
<br> if find(get_dataline(fname, idx), self.match) != -1:
<br> rval = "parse"
<br> else:
<br>! if re.search(self.regexp):
<br> rval = "parse"
<br> if rval != "nothing":
<br> self.trigger_cnt += 1
<br>--- 2331,2337 ----
<br> if find(get_dataline(fname, idx), self.match) != -1:
<br> rval = "parse"
<br> else:
<br>! if re.search(self.regexp, get_dataline(fname, idx)):
<br> rval = "parse"
<br> if rval != "nothing":
<br> self.trigger_cnt += 1
<br><br></tt><hr align="left" width="300" /><br />---------------------------------------------------------------------
<br>To unsubscribe, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470412&i=0" target="_top" rel="nofollow">users-unsubscribe@...</a>
<br>For additional commands, e-mail: <a href="http://old.nabble.com/user/SendEmail.jtp?type=post&post=9470412&i=1" target="_top" rel="nofollow">users-help@...</a><p>From forum: <a href="http://old.nabble.com/perfbase---users-f9271.html" embed="fixTarget[9271]" target="_top" >perfbase - users</a></p>